Escolar Documentos
Profissional Documentos
Cultura Documentos
Cloud Computing
Chandrashekhar S. Pawar
Rajnikant B. Wagh
M. E. student
Department of Computer Engineering
R. C. Patel Institute of Technology
Shirpur, India
pawar.chandrashekhar09@yahoo.com
Associate Professor
Department of Computer Engineering
R. C. Patel Institute of Technology
Shirpur, India
raj_wagh@rediffmail.com
I.
INTRODUCTION
RELATED WORK
USE TECHNIQUES
Input: AP(R,AR)
availableVMList //list of available VMs form each
cloud
usedVMList //list of VMs,currently provision to
certain job
deployableVm=null
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
if size(usedVMList)=size(availbleVMList) then
clear usedVMList
End if
for vm in (AP,R,AR) do
if vm not in usedVMList then
Add VM to usedVMList
deployableVm= vm
Break
End if
End for
15. Return deployableVm
EST ( v i )
LST ( v i )
{ EST ( v m ) AT ( v m )} (1 )
max
v m pred ( v i )
max
{ LST ( v m )} AT ( v i ) ( 2 )
v m succ ( v i )
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
SCHEDULING ALGORITHM
20.
queueTime > waitingTime
21.
Deployed=false
22. End if
23. If deployed then
24. return successful
25. terminate
26. Else
27.
return failure
28.
terminate
As shown above in Algorithm 4, the customers service
deployment requests (R), which is composed of the SLA terms
(S) and the application data (A) to be provisioned, is provided
as input to scheduler (line 1 in Algorithm 1). When service
request (i.e. job) arrive at cloud scheduler, scheduler divide it
in tasks as per there dependencies then the Algorithm 2 is
called to form the list of tasks based to their priority (line 2). In
the first step, it extracts the SLA terms, which forms the basis
for finding the VM with the appropriate resources for
deploying the application. In next step, it collects the
information about the number of running VMs in each cloud
and the total available resources (AR) (line 3). According to
SLA terms appropriate VMs (AP) list is form, which are
capable of provisioning the requested service (R) (lines 4-5).
Once the list of appropriate VMs is form, the Algorithm 1load-balancer decides which particular VM is allocated to
service request in order to balance the load in the data center of
each cloud (lines 6-9).
When there is no VM with the appropriate resources running
in the data center of any cloud, the scheduler checks if service
request (R) has advance reservation then it search for besteffort task running on any cloud or not, if it found best-effort
task then it calls Algorithm 3 CMMS for executing advance
reservation request by preempting best-effort task (lines 1012). If no best-effort task is found on any cloud then scheduler
checks whether the global resources consisting of physical
resources can host new VMs, if yes then, it automatically starts
new VMs with predefined resource capacities to provision
service requests (lines 13-17). But when global resources are
not sufficient to host extra VMs, the provisioning of service
request is place in queue by the scheduler until a VM with
appropriate resources is available (lines 18-22). If after a
certain period of time, the service requests can be scheduled
and deployed, then scheduler returns a scheduling success to
the cloud admin, otherwise it returns failure (lines 23-28).
B. Results
Figure 2 shows the average job execution time in loose
situation. We find out that the PBSA algorithm has the
minimum average execution time. The resource contentions
occur when best-effort job is preempted by AR job. As
resource contention less in loose situation, so that estimated
finish time of job is close to the actual finish time. Hence
adaptive procedure does not impact the job execution time
significantly.
V. EXPERIMENTAL RESULTS
A. Experiment setup
We evaluate the performance of our priority based
scheduling algorithm through simulations. With different
set of jobs simulation is done in 10 runs. In each run of
simulation, we simulate a set of 70 different service
requests (i.e. jobs), and each service request is composed of
up to 18 sub-tasks. We consider 4 clouds in the simulation.
All 70 service requests will be submitted to random clouds
at arbitrary arrival time. Among these 70 service request, 15
requests are in the AR modes, while the rest are in the besteffort modes, with different SLA objectives. The
parameters in Table 1 are set in simulation randomly
according to their maximum and minimum values. Since we
focus only on the scheduling algorithms, we do our
CONCLUSIONS
[11] Jiayin Li, Meikang Qiu, Jian-Wei Niu, Yu Chen, Zhong Ming,
Adaptive Resource Allocation for Preempt able Jobs in Cloud
Systems, in 10th International Conference on Intelligent System
Design and Application, Jan. 2011, pp. 31-36.
[12] Yazir Y.O., Matthews C., Farahbod R., Neville S., Guitouni A.,
Ganti S., Coady Y., Dynamic resource allocation based on
distributed multiple criteria decisions in computing cloud, in 3rd
International Conference on Cloud Computing, Aug. 2010, pp.
91-98.
[13] Goudarzi H., Pedram M., Multi-dimensional SLA-based Resource
Allocation for Multi-tier Cloud Computing Systems, in
IEEE
International Conference on Cloud Computing, Sep. 2011, pp. 324331.
[14] Shi J.Y., Taifi M., Khreishah A.,Resource Planning for Parallel
Processing in the Cloud, in IEEE 13th International Conference on
High Performance and Computing, Nov. 2011, pp. 828-833.
[15] Aoun R., Doumith E.A., Gagnaire M., Resource Provisioning for
Enriched Services in Cloud Environment, IEEE Second
International Conference on Cloud Computing Technology and
Science, Feb. 2011, pp. 296-303.
[16] T. Erl, Service-oriented Architecture: Concepts, Technology, and
Design, Upper Saddle River, Prentice Hall, 2005.
[17] F. Chong, G. Carraro, and R. Wolter,Multi-Tenant Data
Architecture, Microsoft Corporation, 2006.
[18] E. Knorr, Software as a service: The next big thing, InfoWorld,
March 2006.
[19] V. Ungureanu, B. Melamed, and M.Katehakis,Effective Load
Balancing for Cluster-Based Servers Employing Job Preemption,
Performance Evaluation, 65(8), July 2008, pp. 606-622.
[20] L. Aversa and A. Bestavros. Load Balancing a Cluster of Web
Servers using Distributed Packet Rewriting, Proceedings of the 19th
IEEE International Performance, Computing, and Communication
Conference, Phoenix, AZ, Feb. 2000, pp. 24-29.
[21] V. Cardellini, M. Colajanni, P. S. Yu,Dynamic Load Balancing on
Web-Server Systems, IEEE Internet Computing, Vol. 33, May-June
1999, pp. 28 -39.
[22] Chieu T.C., Mohindra A., Karve A.A., Segal A., Dynamic Scaling
of Web Applications in a Virtualized Cloud Computing
Environment, in IEEE International Conference on e-Business
Engineering, Dec. 2009, pp. 281-286.
[23] Naidila Sadashiv, S. M Dilip Kumar, Cluster, Grid and Cloud
Computing: A Detailed Comparison, The 6th International
Conference on Computer Science & Education (ICCSE 2011) August
3-5, 2011. SuperStar Virgo, Singapore, pp. 477- 482.
[24] Vincent C. Emeakaroha, Ivona Brandic, Michael Maurer, Ivan
Breskovic, SLA-Aware Application Deployment and Resource
Allocation in Clouds, 35th IEEE Annual Computer Software and
Application Conference Workshops, 2011, pp. 298-303.
[25] V. C. Emeakaroha, I. Brandic, M. Maurer, and S. Dustdar, Low level
metrics to high level SLAs - LoM2HiS framework: Bridging the gap
between monitored metrics and SLA parameters in cloud
environments, In High Performance Computing and Simulation
Conference, pages 48 55, Caen, France, 2010.
[26]M. Maurer, I. Brandic, V. C. Emeakaroha, and S. Dustdar, Towards
knowledge management in self-adaptable clouds, In 4th International
Workshop of Software Engineering for Adaptive Service-Oriented
Systems (SEASS10) , Miami, Florida, USA, 2010.
[27] T. Hagras and J. Janecek, A high performance, low complexity
algorithm for compile-time task scheduling in heterogeneous
systems, Parallel Computing, vol. 31, no. 7, pp. 653670, 2005.
[28] O. H. Ibarra and C. E. Kim, Heuristic Algorithms for Scheduling
Independent Tasks on Non-identical Processors, Journal of the
ACM, pp. 280289, 1977.